【pyhon 版本 3.5.0 skit-learn版本<0.18.1>】

昨天發現的問題，RandomizedSearchCV怎麼都調不通：

# Split the dataset in two equal parts
X_train, X_test, y_train, y_test = train_test_split(
    data,label, test_size=0.25, random_state=0)
 
# Set the parameters by cross-validation
tuned_parameters = [{'n_neighbors': range(2,7)},
                     {'leaf_size':range(9,100,3)},
                     {'p':range(1,5)}]
 
svr=KNeighborsClassifier()
 
scores = ['precision', 'recall']
 
for score in scores:
    print("# Tuning hyper-parameters for %s" % score)
    print()
 
    labels=y_train.values
    aa
    c, r = labels.shape
    labels = labels.reshape(c,)
 
    clf = RandomizedSearchCV(svr, tuned_parameters,cv=5,n_jobs=-1,verbose=3)
#    clf = GridSearchCV(svr, tuned_parameters,cv=5,n_jobs=-1,verbose=3)
clf.fit(X_train, labels)
 

報錯如下：

 File "C:\ProgramData\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", line 102, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)
 
  File "C:/Users/gzhuangzhongyi/Desktop/NetEase/test/RandomSearchCV_Functional.py", line 46, in <module>
    clf.fit(X_train, labels)
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py", line 1190, in fit
    return self._fit(X, y, groups, sampled_params)
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py", line 564, in _fit
    for parameters in parameter_iterable
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\externals\joblib\parallel.py", line 758, in __call__
    while self.dispatch_one_batch(iterator):
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\externals\joblib\parallel.py", line 603, in dispatch_one_batch
    tasks = BatchedCalls(itertools.islice(iterator, batch_size))
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\externals\joblib\parallel.py", line 127, in __init__
    self.items = list(iterator_slice)
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py", line 557, in <genexpr>
    )(delayed(_fit_and_score)(clone(base_estimator), X, y, self.scorer_,
 
  File "C:\ProgramData\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py", line 230, in __iter__
    for v in self.param_distributions.values()])
 
AttributeError: 'list' object has no attribute 'values'
 

經過檢視fit方法，發現無論如何調整fit方法的引數，都沒法執行。

但是如果換成GridSearchCV就可以執行。

經過檢視類實現，發現兩種類呼叫了相同的，fit方法，但是，fit方法有隱含傳入的引數：

sampled_params = ParameterSampler(self.param_distributions,
                                          self.n_iter,
                                          random_state=self.random_state)
        return self._fit(X, y, groups, sampled_params)
 

其中，sampled_params為傳入引數之取樣。

其傳入引數在初始化的時候傳入：

clf = RandomizedSearchCV(svr, tuned_parameters,cv=5,n_jobs=-1,verbose=3)

而，這個引數由：

tuned_parameters = [{'n_neighbors': range(2,7)},
                     {'leaf_size':range(9,100,3)},
                     {'p':range(1,5)}]

語句設定，這裡有三個字典。而正確的是：

tuned_parameters = [{'n_neighbors': range(2,7),
                     'leaf_size':range(9,100,3),
                     'p':range(1,5)}]

Grid的時候會遍歷字典中所有引數的組合，所以字典的劃分不重要。

 for p in self.param_grid:
            # Always sort the keys of a dictionary, for reproducibility
            items = sorted(p.items())
            if not items:
                yield {}
            else:
                keys, values = zip(*items)
                for v in product(*values):
                    params = dict(zip(keys, v))
                    yield params

但是Randomlize,當傳入字典的時候，會作為帶分佈的進行處理，對字典取值

# Always sort the keys of a dictionary, for reproducibility
            items = sorted(self.param_distributions.items())
            for _ in six.moves.range(self.n_iter):
                params = dict()
                for k, v in items:
                    if hasattr(v, "rvs"):
                        if sp_version < (0, 16):
                            params[k] = v.rvs()
                        else:
                            params[k] = v.rvs(random_state=rnd)
                    else:
                        params[k] = v[rnd.randint(len(v))]
                yield params

Random會檢查傳入的引數，如果可以遍歷就認為是分佈。

於是傳入作為fit的引數集的時候，不是作為可遍歷的物件的字典，可以.values，而是一個一個把分佈元素組合成字典的list，但因為傳入的不是一個分佈而是一個list，所以不能對分佈取值。

上面的兩段函式GridSearchCV產生的引數集：

RandomizeSearchCV產生的引數集因為debug調不出來，無法展示。